SK hynix unveils 'iHBM' thermal architecture that cools AI memory at the source — integrated cooling elements inside HBM interface cut thermal resistance by 30%, target next-gen HBM5 accelerators and dense AI data centers
SK hynix has introduced 'iHBM,' a novel thermal architecture designed to address the critical heating issues in High-Bandwidth Memory (HBM) used for AI acceleration. This technology integrates cooling elements directly into the memory interface, reducing thermal resistance by over 30%. The innovation aims to prevent thermal throttling, enabling future memory generations like HBM5 to operate at higher capacities and speeds within dense AI data centers. AI
IMPACT This innovation could alleviate thermal bottlenecks, enabling more powerful and efficient AI hardware by allowing memory to operate at higher speeds and densities.